Skip to content

AI Content Detector

Tighten your writing before you publish it.

Writing ToolsFreeRuns in your browser
Interactive ai content detectorRuns in your browser, nothing uploadedFree, no account
Overview

What is the AI Content Detector?

No AI content detector — including this one — can reliably prove whether a specific piece of text was written by AI. This tool does not attempt to. Instead it measures four statistical writing-pattern signals directly from your text: sentence-length variance (burstiness), vocabulary diversity, repetition of transition phrases over-represented in AI-assistant writing, and punctuation-pattern uniformity. Each is a genuine, computable statistic — a standard deviation, a ratio, a phrase count — shown with the exact method used to calculate it. None of them, alone or combined, is evidence that a person did or did not write something, and this tool never outputs a single "% AI-written" score, because that number would be fabricated confidence dressed up as a measurement.

The most important limitation to understand before using this tool: false positives disproportionately affect non-native English writers. Sentence-length variance and vocabulary diversity both tend to be lower in the writing of people composing in a second language, for reasons that have nothing to do with whether AI was involved — simpler sentence structures and a narrower working vocabulary are a normal, well-documented feature of non-native writing at every skill level. A detector that flags "low burstiness, low vocabulary diversity" as suspicious will systematically flag non-native writers more often than native ones, regardless of who actually wrote the text. Never use this tool's output to accuse a student, colleague or contributor of using AI — it is not evidence, and treating it as evidence causes real harm to real people.

An honest note on the data

No AI detector is reliable enough to accuse anyone. This tool reports statistical signals (burstiness, perplexity proxies) as indicators only.

Benefits

Why use a free ai content detector?

Four named, measurable signals — not a black-box score

Burstiness, vocabulary diversity, transition-phrase density and punctuation uniformity are each shown with the exact statistic and formula behind them, so you can see precisely what was measured.

No fabricated "% AI-written" number

The tool deliberately does not output a single probability or percentage, because no defensible calculation for that number exists without ground truth this tool cannot have.

Explains the false-positive risk up front

The tool states plainly, before you even see a result, that these signals systematically misfire on non-native English writing — a limitation many detectors do not disclose.

Shows the sentence-length pattern visually

A bar chart of every sentence's word count makes burstiness (or the lack of it) visible at a glance, rather than reducing it to an opaque single number.

Runs entirely in your browser

All four calculations happen locally — nothing you paste is sent to a server or an AI model to be analysed.

Guide

How do you use the AI Content Detector?

  1. 1

    Read the warning first

    Before pasting anything, understand what these signals can and cannot tell you — they are not proof of authorship, and false positives are common, especially for non-native English writers.

  2. 2

    Paste the text to analyse

    Results update automatically. Longer text (several paragraphs) gives more statistically meaningful sentence-length and vocabulary measurements than a single sentence.

  3. 3

    Review each of the four signals individually

    Each panel shows the measured value, a low/medium/high label, and the exact method used — read the method, not just the label, before drawing any conclusion.

  4. 4

    Treat the result as a prompt for discussion, not a verdict

    If a signal looks unusual, the appropriate response is a conversation with the writer about their process, not an accusation.

Accuracy

How accurate is it?

Every number shown is a directly computed statistic from your text, not a model prediction: burstiness is the coefficient of variation of sentence length (standard deviation ÷ mean), vocabulary diversity is a type-token ratio (unique words ÷ total words), transition-phrase density is a literal count of specific phrases per 100 words, and punctuation uniformity is the coefficient of variation of commas per sentence. All four are calculable with certainty. What is not certain — and cannot be made certain by any tool — is what any combination of these numbers means about who or what wrote the text.

  • Burstiness: standard deviation of sentence length ÷ mean sentence length, across every sentence in the text. Human writing that mixes short, punchy sentences with long, complex ones scores higher; uniform sentence length scores lower. Careful human editors who write consistently will also score low here.
  • Vocabulary diversity: unique words ÷ total words (type-token ratio). A technical piece that deliberately repeats a key term will score low on this measure for entirely legitimate reasons.
  • Transition-phrase density: a count of phrases over-represented in AI-assistant writing ("moreover", "in conclusion", "it is important to note", "plays a crucial role" and similar), normalised per 100 words. Formal academic and business writing also uses many of these phrases naturally.
  • Punctuation uniformity: coefficient of variation of commas per sentence, plus how many distinct terminal marks (., !, ?) appear across the whole text. A narrow, repetitive punctuation rhythm is measured, not explained.
  • None of these four signals were validated against a labelled dataset of confirmed AI-written and confirmed human-written text by this tool — they are general, published stylometric measures, applied honestly and shown with their exact method, not claimed as a trained or calibrated detection model.
Details

What should you know before using it?

What each signal label means, and how much weight it should carry:

SignalWhat a "low" or "high" reading actually indicates
Burstiness (low)Sentence lengths are unusually uniform. Common in some AI writing, but also in formal, tightly edited or non-native English prose.
Vocabulary diversity (low)A narrower range of distinct words relative to text length. Common in technical writing, short texts, and non-native English writing.
Transition-phrase density (high)Frequent use of phrases like "moreover" or "in conclusion". Also common in formal academic and business writing styles taught explicitly in many non-native English curricula.
Punctuation uniformity (high)A consistent comma-per-sentence pattern and narrow use of terminal punctuation. Also typical of careful, rule-following editing.
Applications

When should you use a ai content detector?

Because the ai content detector runs entirely in your browser, it suits work you would not want to hand to a third-party server — client files, unpublished drafts, anything under an NDA. These are the situations people reach for it in most often.

Reviewing your own writing before publishing

Check whether your own draft leans heavily on flagged transition phrases or reads with unusually uniform sentence rhythm — useful for improving style, independent of any authorship question.

Starting an informed conversation, not an accusation

If a piece of writing looks unusual, use these signals as a prompt to ask the writer about their process — never as standalone grounds for a disciplinary or academic-integrity decision.

Editing AI-assisted drafts for a more natural rhythm

If you used an AI tool as a starting point and want the final piece to read less uniformly, the burstiness and transition-phrase panels show you concretely where to vary sentence length and wording.

Understanding detector limitations before relying on any tool

Use this page's explanation of false-positive risk as a reference before trusting the output of any AI-detection tool, including commercial ones that present a confident-looking percentage.

FAQs

Frequently asked questions

Can this tool prove whether text was written by AI?
No. No AI detector, including this one, can reliably prove authorship. This tool measures four statistical writing-pattern signals and shows you the exact method behind each one, but it never outputs a probability or verdict, because no honest calculation for that number exists.
Why doesn't it give me a percentage, like other AI detectors do?
Because that percentage would not be a real measurement — it would be a number presented with false confidence. This tool shows four separate, genuinely computed statistics instead (sentence-length variance, vocabulary diversity, transition-phrase density, punctuation uniformity) so you can see exactly what was measured, rather than trusting an opaque score.
Does this tool discriminate against non-native English writers?
The underlying signals do carry that risk, and this page states it directly: sentence-length variance and vocabulary diversity both tend to run lower in non-native English writing for reasons unrelated to AI use. This tool discloses that risk prominently instead of hiding it, but the risk exists in any tool using these same statistical measures — that is precisely why no result here should be treated as evidence against a specific person.
What is "burstiness" and why does it matter?
Burstiness measures how much sentence length varies across a piece of text — the coefficient of variation of words-per-sentence. Writing that mixes short and long sentences scores higher; writing with uniform sentence length scores lower. Some AI-generated text shows lower burstiness, but so does plenty of carefully edited human writing, so a low score is a data point, not a conclusion.
What are the "transition phrases" it looks for?
A list of phrases that show up disproportionately often in text produced by AI writing assistants — things like "moreover", "in conclusion", "it is important to note" and "plays a crucial role". The tool counts how often they appear per 100 words. These phrases are also standard in formal academic and business writing, so frequent use alone proves nothing.
Should I use this to grade student work or accuse someone of using AI?
No. This tool explicitly warns against that use. Statistical writing-pattern signals are not evidence of authorship, and using them to make an accusation — especially against a non-native English speaker, who is more likely to trigger a false positive — causes real harm based on an unreliable signal.
Is my text sent to an AI model to be analysed?
No. All four calculations (standard deviation, ratios, phrase counts) run directly in your browser using JavaScript. Nothing you paste is sent to a server or a language model.
Security

Is the ai content detector safe to use?

This tool runs entirely inside your browser. Your text, files and settings are processed on your own device and are never uploaded to our servers — there is nothing for us to store, log or leak. You can confirm it yourself: open your browser DevTools, switch to the Network tab, and use the tool. You will not see an upload request. It also means the tool keeps working if your connection drops mid-task.

No account required No file storage HTTPS everywhere
Get in touch

Need something this tool can’t do?

If you need a feature added, spotted something wrong, or want a custom tool or website built for your business, tell us. We read every message and we build what people actually ask for.